Impacto de datos sintéticos generados mediante CTGAN en modelos de aprendizaje automático para la predicción
DOI:
https://doi.org/10.35381/i.p.v8i15.5249Palabras clave:
Inteligencia artificial, tratamiento de datos, análisis de datos, estadística, aprendizaje en línea, (Tesauro UNESCO)Resumen
Este estudio tuvo como objetivo evaluar comparativamente el desempeño predictivo y la explicabilidad de Regresión Logística, Random Forest y XGBoost entrenados con datos reales frente a datos reales aumentados con CTGAN. La presente investigación es de naturaleza computacional/experimental sobre modelos, de nivel predictivo, con enfoque cuantitativo, propósito comparativo y orientación aplicada. Los resultados expresaron aspectos contrarios a lo que reporta la mayor parte de la literatura consultada, añadir datos sintéticos al entrenamiento no mejoró el desempeño predictivo de ninguno de los tres modelos, sino que lo redujo de forma estadísticamente significativa, incluso cuando esos mismos datos sintéticos superaron la validación estadística de fidelidad frente a los datos reales. En conclusión, a nivel teórico, el estudio aporta un caso concreto sobre la relación entre fidelidad estadística y utilidad predictiva de los datos sintéticos, un aspecto todavía poco explorado en la literatura revisada.
Descargas
Citas
Adiputra, I. N. M., Lin, P.-C., & Wanchai, P. (2025). The effectiveness of generative adversarial network-based oversampling methods for imbalanced multi-class credit score classification. Electronics, 14(4), Article 697. https://doi.org/10.3390/electronics14040697
Ali Shahee, S., & Patel, R. (2025). An explainable ADASYN-based focal loss approach for credit assessment. Journal of Forecasting, 44(4), 1513–1530. https://doi.org/10.1002/for.3252
Borisov, V., Leemann, T., Seßler, K., Haug, J., Pawelczyk, M., & Kasneci, G. (2024). Deep neural networks and tabular data: A survey. IEEE Transactions on Neural Networks and Learning Systems, 35(6), 7499–7519. https://n9.cl/khip9
Breiman, L. (2001). Random forests. Machine Learning, 45(1), 5–32. https://doi.org/10.1023/A:1010933404324
Chen, T., & Guestrin, C. (2016). XGBoost: A scalable tree boosting system. En Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (pp. 785–794). https://doi.org/10.1145/2939672.2939785
El Sayad, I., Chong, F. H. L., & Gosal, B. (2026). Risk assessment in mortgages: A comparative study of AI models. PeerJ Computer Science, 12, Article e3494. https://doi.org/10.7717/peerj-cs.3494
Gao, X., & Zhou, Y. (2026). A resampling ensemble model for multi-window corporate default prediction under class imbalance. Systems, 14(7), Article 776. https://n9.cl/nk6n55
Gómez-de-Mariscal, E., Guerrero, V., Sneider, A., Jayatilaka, H., Phillip, J. M., Wirtz, D., & Muñoz-Barrutia, A. (2021). Use of the p-values as a size-dependent function to address practical differences when analyzing large datasets. Scientific Reports, 11, Article 20942. https://n9.cl/je6yl
Kheddar, H., Himeur, Y., Al-Maadeed, S., Amira, A., & Bensaali, F. (2023). Deep transfer learning for automatic speech recognition: Towards better generalization. Knowledge-Based Systems, 277, Article 110851. https://doi.org/10.1016/j.knosys.2023.110851
Li, Z., Chen, G., Liu, Y., & Yuan, J. (2025). An efficient multi-institution collaborative learning framework based on knowledge filtering and data sharing. Cluster Computing, 28(10), Article 644. https://n9.cl/wmjvdd
Lundberg, S. M., & Lee, S.-I. (2017). A unified approach to interpreting model predictions. Advances in Neural Information Processing Systems, 30. https://n9.cl/82snlp
Petersen, K., Feldt, R., Mujtaba, S., & Mattsson, M. (2008). Systematic mapping studies in software engineering. En 12th International Conference on Evaluation and Assessment in Software Engineering (EASE). BCS Learning & Development. https://doi.org/10.14236/ewic/EASE2008.8
Petticrew, M., & Roberts, H. (2006). Systematic reviews in the social sciences: A practical guide. Blackwell Publishing. https://n9.cl/rw7s1r
Pourkhoshgoftar, S., Shahbahrami, A., & Esmi, N. (2025). Graph-based inductive learning for credit risk prediction with imbalance mitigation. Computational Economics. https://doi.org/10.1007/s10614-025-11114-9
Raheem, E. A., Dinar, A. M., Mohammed, M. A., & Al-Attar, B. (2024). Improving loan status prediction accuracy with generative adversarial networks: Addressing data scarcity and bias. Journal of Intelligent Systems and Internet of Things, 13(1), 251–258. https://doi.org/10.54216/JISIoT.130118
Ramachandra, P., & Vaithiyanathan, S. (2025). Fed-DPSDG-WGAN: Differentially private synthetic data generation for loan default prediction via federated Wasserstein GAN. IEEE Access, 13, 52069–52084. https://n9.cl/knhnf
Shan, W., Kang, X., & Gao, B. (2026). A TabPFN-based framework for credit risk prediction in automotive green supply chain finance. Sustainability, 18(12), Article 6305. https://doi.org/10.3390/su18126305
Singhal, P., Walambe, R., Ramanna, S., & Kotecha, K. (2023). Domain adaptation: Challenges, methods, datasets, and applications. IEEE Access, 11, 6973–7020. https://n9.cl/3wvi76
van Eck, N. J., & Waltman, L. (2010). Software survey: VOSviewer, a computer program for bibliometric mapping. Scientometrics, 84(2), 523–538. https://doi.org/10.1007/s11192-009-0146-3
Xu, L., Skoularidou, M., Cuesta-Infante, A., & Veeramachaneni, K. (2019). Modeling tabular data using conditional GAN. Advances in Neural Information Processing Systems, 32. https://n9.cl/rfe5on
Xueping, Z., Samuri, S. M., & Adnan, M. H. M. (2026). Adaptive anomaly correction and dynamic sample balancing for deep intelligent credit risk assessment framework. NED University Journal of Research, 23(1), 44–69. https://doi.org/10.35453/NEDJR-ASCN-2025-007.R4
Zhao, X., & Guan, S. (2023). CTCN: A novel credit card fraud detection method based on Conditional Tabular Generative Adversarial Networks and Temporal Convolutional Network. PeerJ Computer Science, 9, Article e1634. https://doi.org/10.7717/peerj-cs.1634
Publicado
Cómo citar
Número
Sección
Licencia
Derechos de autor 2026 Wilmer Illescas-Espinoza, Nancy Loja-Mora, Wilmer Rivas-Asanza
Esta obra está bajo una licencia internacional Creative Commons Atribución-NoComercial-CompartirIgual 4.0.
CC BY-NC-SA : Esta licencia permite a los reutilizadores distribuir, remezclar, adaptar y construir sobre el material en cualquier medio o formato solo con fines no comerciales, y solo siempre y cuando se dé la atribución al creador. Si remezcla, adapta o construye sobre el material, debe licenciar el material modificado bajo términos idénticos.
OAI-PMH URL: https://fundacionkoinonia.com.ve/ojs/index.php/ingeniumetpotentia/oai
